# Reported issues for Mnemoverse Memory Pod holds 6 of 6 GitHub reports that passed its relevance review. This can include external user reports, maintainer-confirmed bugs, and concrete feature gaps. Treat them as evidence to inspect, not a count of distinct defects. Back to [Mnemoverse Memory](/mcp/mnemoverse-memory). ## Most discussed ### Remote MCP: writes can succeed while every read returns "not connected" — and the client still reports Connected (local stdio works) > **UPDATE 2026-09-05 — the diagnosis in this issue is now too narrow; see the reproduction in the comments.** > > The title said "the session dies". A fresh run shows something more specific and more actionable: in one session the remote connector **wrote successfully** to a shared room and then could not **read** anything for ninety minutes, while `claude mcp list` reported it `Connected` throughout. In the same session, same account, same API key, the **local stdio server… [Read the thread](https://github.com/mnemoverse/mcp-memory-server/issues/97) · 2026-08-22 · open · 4 comments ### Release did not complete: v0.12.0 Release did not complete for **v0.12.0**. The release workflow did not complete — it failed or was cancelled — so at least one public surface is not serving this version. [Open the run](https://github.com/mnemoverse/mcp-memory-server/actions/runs/36117440684). **Read the failing step before assuming.** The one that has bitten us: > `npm error code E404` — `PUT https://registry.npmjs.org/@mnemoverse%2fmcp-memory-server` A 404 on `PUT` for a package that plainly exists is **not** "package not… [Read the thread](https://github.com/mnemoverse/mcp-memory-server/issues/173) · 2026-09-25 · closed · external user · 2 comments ### Post-0.9.1: memory_list_recent vanished from a remote consumer's tool surface; a second session lost the whole server — same day ## Two independent tool-surface degradations observed on 2026-08-24, hours after the v0.9.1 release Facts only; needs server-side log correlation. Source at HEAD still ships `memory_list_recent` (5 references in `src/index.ts`), so this is not a code removal — it looks like a serving/deploy-surface incident. **Observation A (remote/HTTP consumer, ~19:5x UTC).** An agent that had been using the server for weeks reported that after "a tool surface update" `memory_list_recent` disappeared from… [Read the thread](https://github.com/mnemoverse/mcp-memory-server/issues/113) · 2026-08-24 · closed · 2 comments ### Multilingual rendering: [by …] author tags erase non-Latin names to nothing (policy decision needed) ## What `safeInline` (src/render.ts) maps every character outside `[\w .@:+/-]` to a space. In JavaScript regexes `\w` is ASCII-only, so the sanitiser is de-facto **Latin-only**: a Cyrillic, CJK or Arabic `agent_name` sanitises to an empty string and the ` [by …]` author tag silently disappears from the line (`formatAuthorTag` returns `""` when the sanitised name is empty). 0.8.1 already moved **room names** off this sanitiser onto `exactLiteral` (src/names.ts) — readable non-ASCII,… [Read the thread](https://github.com/mnemoverse/mcp-memory-server/issues/66) · 2026-08-08 · closed · 2 comments ### Release did not complete: v0.10.2 Release did not complete for **v0.10.2**. The release workflow did not complete — it failed or was cancelled — so at least one public surface is not serving this version. [Open the run](https://github.com/mnemoverse/mcp-memory-server/actions/runs/35517465176). **Read the failing step before assuming.** The one that has bitten us: > `npm error code E404` — `PUT https://registry.npmjs.org/@mnemoverse%2fmcp-memory-server` A 404 on `PUT` for a package that plainly exists is **not** "package not… [Read the thread](https://github.com/mnemoverse/mcp-memory-server/issues/140) · 2026-09-20 · closed · external user · 1 comment ### Public surfaces are not tracking the released version `release-sync check` is red. At least one first-party surface (npm, the Official MCP Registry, the GitHub release) either did **not match** `package.json#version`, or could **not be checked at all**. Those are different events and the run log reports them separately, surface by surface. [Open the run](https://github.com/mnemoverse/mcp-memory-server/actions/runs/35428417654). Read the log before acting, because the two outcomes call for opposite responses: * `✗ DRIFT (have vX.Y.Z)` means the… [Read the thread](https://github.com/mnemoverse/mcp-memory-server/issues/134) · 2026-09-19 · closed · external user · 1 comment ## Most recent The remaining reports are on [the project's issue tracker](https://github.com/mnemoverse/mcp-memory-server/issues).